Representativity and Networked Interference in Data-Rich Field Experiments - A Large-Scale RCT in Rural Mexico

Jul 19, 2026

Modern availability of rich geospatial datasets and analytic tools enables more informed design of field experiments, especially when sampling choices affect external representativity and networked interference, interpreted as SUTVA violations. The paper proposes a two-part methodology: modeling geospatial and social interaction drivers of interference and generating nondominated sample options approximating the Pareto tradeoff between interference and representativity. The approach is developed and tested in a large rural Mexico health study with 3,000+ pregnant women and 600 clinics.

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages5

Reading Time13 min

Abstract

A Large-Scale Digital Health Experiment in Rural Mexico

Prospera and the digital communication interventions

External Representativity and Interference in Sample Choice

Population average treatment effect and extrapolation

Modeling Interference Networks Using Geospatial Data and GIS Tools

Interference gravity model

Pub lic Disclosure Authorized Pub lic Disclosure Authorized The World Bank Economic Review, 34(Supplement), 2020, S35–S39 doi: 10.1093/wber/lhz038 Supplementary Article Representativity and Networked Interference in Data-Rich Field Experiments: A Large-Scale RCTin Rural Mexico Alejandro Noriega and Alex Pentland Abstract Modern availability of rich geospatial datasets and analysis tools can provide insight germane to the design of field experiments. Design of field experiments, and in particular the choice of sampling strategy, requires careful consideration of its consequences on the external representativity and interference (SUTVA violations) of the experimental sample. This paper presents a methodology for a) modeling the geospatial and social interaction factors that drive interference in rural field experiments; and b) eliciting a set of nondominated sample options that approximate the Pareto-optimal tradeoff between interference and external representativity, as functions of sample choice. The study develops and tests the methodology in the context of a large-scale health experiment in rural Mexico, involving more than 3,000 pregnant women and 600 health clinics across 5 states. Relevant for the practitioner, the methodology is computationally tractable and can be implemented leveraging open sourced geo-spatial data and software tools. JEL classification: I00, C1, C6, C8, C9 Keywords: field experiments, networked interference, SUTVA violations, big data, spatial analysis, cash transfers 1. A Large-Scale Digital Health Experiment in Rural Mexico Mexico’s social assistance program Prospera is one of the largest conditional cash programs in the world, providing health care to nearly 30 million beneficiaries (Oportunidades 2011) . The physical remoteness of Prospera’s rural beneficiaries drives key challenges in provisioning its services. Moreover, information in Prospera flows through traditional means such as fliers, radio announcements, and door-to-door communication. National authorities have endeavored to introduce digital means of communication with, and among, Prospera beneficiaries.1 In this context, this study participated in designing a large-scale randomized control trial (RCT) to assess the effect of such potential interventions on health outcomes. The experiment focuses on maternal and child health; involves more than 600 health clinics and 3,000 pregnant women Alejandro Noriega (corresponding author) is a visiting scholar at the MIT Human Dynamics Laboratory, and founder of Prosperia Labs; his email address is [anc@prosperia.ai](anc@prosperia.ai). Alex Pentland is director of the MIT Human Dynamics Laboratory, Academic Director of the Harvard-MIT-ODI DataPop Alliance, and Faculty Director of theMIT Connection Science Research Initiative, Cambridge, MA; his email address [is pentland@mit.edu](is pentland@mit.edu). 1 In partnership with a set of academic institutions and NGOs, such as the MIT Media Laboratories. © The Author(s) 2019 . Published by Oxford University Press on behalf of the International Bank for Reconstruction and Development / THE WORLD BANK. All rights reserved. For permissions, please e-mail: [journals.permissions@oup.com](journals.permissions@oup.com) Downloaded from [https://academic.oup.com/wber/article/34/Supplement_1/S35/5686213 by guest on 23 June 2021](https://academic.oup.com/wber/article/34/Supplement_1/S35/5686213 by guest on 23 June 2021) across 5 states; and tests 3 treatment arms, consisting of top-down, peer-to-peer, and down-to-top feedback communication. 2. External Representativity and Interference in Sample Choice This paper focuses on extrapolation in the common setting where random sampling is not viable in practice, such as in the Prospera experiment. As pointed out by Muller (2015), sampling at random is commonly “not practically feasible, or researchers have the more ambitious aim of generalizing beyond a single, prespecified population

FAQs

Why does sampling strategy matter in data-rich field experiments?

Sampling affects external representativity and can induce interference that violates SUTVA assumptions. The paper emphasizes that consequences of sample choice must be carefully considered.

What methodology does the paper propose for dealing with representativity-interference tradeoffs?

It models geospatial and social interaction factors driving interference and elicits nondominated sample options that approximate the Pareto-optimal tradeoff between interference and representativity.

How is the proposed methodology tested in the study?

It is applied to a large-scale randomized health experiment in rural Mexico involving more than 3,000 pregnant women and 600 health clinics across five states, evaluating communication-based treatment arms.

Jul 19, 2026

Upload by Theodora